Repository navigation
research(zfae): construct gonol input parsing and preserve source evidence - #52
Conversation
…ence Add a source-bound construction track and frozen parser distinction probe. Record six input collisions with passing controls without claiming inference capability or transferring upstream authority. Update Stack projections as one structural transaction.
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
erinepshovel-code
changed the title
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 7398cdb55d
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 3194eee189
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| if not isinstance(gonol, self._gonol_type): | ||
| raise GonolAdmissionError(f"foreign gonol type at occurrence {ordinal}") |
There was a problem hiding this comment.
When load_parser() is called twice, _verified_module re-executes the producer and creates a distinct GlyphGonol class each time, so this isinstance check rejects a gonol produced by the first parser when passed to the second—even though both parsers verified the same declared source blob. This also affects callers that constructed gonols from an exact producer import before loading the parser; validate the source-bound construction without requiring identity with one ephemeral class object.
AGENTS.md reference: AGENTS.md:L28-L30
Useful? React with 👍 / 👎.
1 participant
ZFAE needs a constructed input boundary before neural inference can be built. This PR starts that research in Stack and supplies a gonol-backed parser using the existing English
GlyphGonolproducer and the exact UCNS Public Gonol carrier.The first experiment found six collisions when A0-betatest's complete heuristic
SemanticFeaturesrecord was proposed as the sole lossless input boundary. The original FALSIFIED receipt is preserved. The new parser distinguishes all six pairs in the native gonol sequences themselves, without relying on source IDs or hashes to create differences.The parser accepts text, strict UTF-8, or ordered native glyph gonols. Repeated glyphs share one immutable construction while retaining separate occurrences and exact scalar/byte spans. Serialized results contain complete native construction records; replay rejects changed identities, spans, construction parts, and provenance. Unsupported scalars remain visible in their original positions, and complete-gonol consumption refuses partial admission.
The declared synthetic profile contains all 157 UCNS carrier positions plus eight explicit non-carrier scalars: NUL, TAB, CR, LF, é, combining acute accent, 中, and 😀. Extras consume the existing English encoded-name/codepoint construction; they do not define UCNS function operations. This is a finite research profile, independent of the OEWN inventory. No second gonol schema or geometry law is introduced.
research/zfae/gonol_parser.py,GONOL_PARSER.jsongonol_probe.py,tests/test_gonol_parser.py, v1 receipt.github/workflows/zfae-research.ymlLocal validation on tree
8402e61509f5cde3826073d40e6da36875036a42:git diff --checkpass. No training, paid API calls, or GPU compute.Exact implementation sources:
d713cc5f901d74deb902a191837ef5657dbaba94, blob8b55823805c87ad8c4cc9d7451dc2eb90bdedd3a.4086ab82399c4d142b0eacfbc09e0a69ed151aa5, blobc1955e46e2dc918fb657cb346e42106d71937e91.abd259b4722901317e4388d774a20d6819d959c2, including gonol-build.70787567cb37f4b156a66b293f9bd9cb5cf7d722f1e6ff1721f5df343ada5b79.Usage is in
research/zfae/README.md. Python 3.12 / Unicode 15.0.0 is required for exact construction replay. The parser is the Stack research input boundary; it does not alter A0-betatest's live decoder, construct higher gonols, or satisfy PTCNA's neural-audit gate. ZFAE retains conceptual authority, UCNS geometry, PTCNA neural construction and a0 runtime integration. UCHC release/reconsumption remains incomplete.Rollback: revert the parser transaction and associated graph/receipt projections; retain sealed historical evidence. No upstream source or
libs/view is modified.hmmm: expand admission only through declared producer profiles; derive higher closure and geometric function operations; bind the UCNS neural audit required by PTCNA; construct and validate inference before runtime integration.